GEORGE C. PRATT, Circuit Judge:
The question presented in this appeal by the United States from the summary judgment in Manufacturers Hanover Trust Co. v. United States, 576 F.Supp. 837 (S.D.N.Y.1983), Charles E. Stewart, Judge, is whether the Internal Revenue Service’s use of gender-based mortality tables to value reversionary interests of a decedent’s estate under 26 U.S.C. § 2037 violates the equal protection guaranteed by the due process clause of the fifth amendment. The IRS, following regulations in effect from 1970 to 1983, used gender-based mortality tables to calculate the estate tax owed by plaintiff Manufacturers Hanover Trust Company, as executor of the estate of Charlotte C. Wallace. The district court held that the IRS practice was unconstitutional, and it therefore awarded plaintiff a tax refund of $455,581.71 plus interest.
We reverse. Although the challenged IRS practice did distinguish between males and females, the gender classification was substantially related to the important governmental objective of promoting equity and fairness in estate taxes by accurately valuing reversionary interests. The challenged practice lacks those characteristics that make gender classifications invidious. There is no evidence that the challenged practice distributes burdens or benefits in a way that disadvantages the class of women as a whole, or that disadvantages the class of men as a whole. The gender classification does not demean the ability or social status either of women or of men, and it is not based on assumptions that women or men will choose stereotyped or traditional social roles. The gender-based mortality tables realistically reflect the fact that men and women have different average life expectancies, and the government’s use of these averages for determining values in estate taxation does not create an unacceptable risk of discriminating against those who are not within the statistical norm. In light of all these circumstances, there is nothing unconstitutional about the challenged practice.
Background
On November 5, 1923, Charlotte C. Wallace established a trust that provided income to her for her life and, on her death, income to her son, Howard. The corpus of the trust would pass by the terms of Howard’s will if he outlived his mother, but by the terms of Charlotte’s will if he predeceased her. Charlotte died on February 28, 1976, at the age of 88; she was survived by Howard, who was then 57 years old. The estate tax law provides that if the value of a decedent’s reversionary interest in a trust immediately before death exceeds 5 percent of the trust’s value, then the trust corpus must be included in the gross estate. 26 U.S.C. § 2037(a)(2) (1982).
Charlotte’s executor calculated the value of her reversionary interest at 4.9867 percent, using Unisex Table 1 of the United States Life Tables: 1959-61, published by what was then known as the United States Department of Health, Education and Welfare. Because the value fell below the statutory 5 percent cutoff, the estate paid no tax on the trust corpus.
[462]*462The Treasury regulations in effect at the time of Charlotte’s death, however, required the value of a reversionary interest in a trust to be computed according to gender-based mortality tables. See Treas. Reg. §§ 20.2031-7(a), 20.2031-10(d), (e), (f), 20.2037-l(c)(3); Rev.Rul. 66-307, 1966-2 Cum.Bull. 429. Using gender-based tables, IRS calculated the value of the reversion-ary interest at 6.654 percent and assessed the estate with an aggregate deficiency, plus additions, of $458,662.98.
The estate paid the deficiency and, after an administrative claim for a refund was denied, filed this refund suit. The district court granted the estate’s motion for summary judgment, 576 F.Supp. 837 (S.D.N.Y.1983), holding that use of gender-based mortality tables violated the equal protection component of the fifth amendment. This appeal followed.
Discussion
“Statutory classifications that distinguish between males and females are ‘subject to scrutiny under the Equal Protection Clause.’ ” Craig v. Boren, 429 U.S. 190, 197, 97 S.Ct. 451, 456, 50 L.Ed.2d 397 (1976) (quoting Reed v. Reed, 404 U.S. 71, 75, 92 S.Ct. 251, 253, 30 L.Ed.2d 225 (1971)). “To withstand constitutional challenge * * * classifications by gender must serve important governmental objectives and must be substantially related to achievement of these objectives.” Craig v. Boren, 429 U.S. at 197, 97 S.Ct. at 456. In the present case the government argues, first, that the challenged statutory scheme does not treat men and women differently, and second, that even if the statutory scheme does treat men and women differently, the differences in treatment are justified and not invidious. We find the arguments supporting the first point to be unpersuasive, but on the second point we agree with the government that there is nothing invidious or unjustified in the way this particular statutory scheme makes use of distinctions between men and women.
A. Are similarly situated men and women treated differently because of classification by gender?
The government argues that the IRS practice does not treat men and women differently because “[ujnder the 5 percent rule of section 2037, the estates of decedents of both sexes are includable, and trusts established for the benefit of members of both sexes are taxable. The key is not the grantor’s or the grantee’s sex, but the value of the grantor’s reversionary interest.”
This argument is unconvincing. While it is true that estates of both male and female decédents are includable under the rule, and that trusts may be taxed whether their beneficiaries are men or women, nevertheless, what the government calls the “key” of section 2037 — the value of the grantor’s reversionary interest — is determined by calculations that treat similarly situated men and women differently. As a result the amount of the estate tax will sometimes depend upon whether the decedent or beneficiary is male or female. This comes about in the following way.
For purposes of the estate tax, the value of the grantor’s reversionary interest in a trust requires calculating the probability that a person at the age of the grantor, just prior to death, would outlive the trust’s beneficiary. The figures on life expectancies come from standard mortality tables, and gender-based tables reflect the undisputed fact that, on the average, women live longer than men at every age. This statistical fact then affects the calculation of a reversionary interest in two ways.
First, when gender-based tables are used, the likelihood that a female trust settlor will outlive her beneficiary is greater than the likelihood that a male trust settlor in similar circumstances will outlive his beneficiary. (Similar circumstances, for present purposes, occur when the male trust settlor’s beneficiary is the same age and gender as the female trust settlor’s beneficiary.) Therefore, when gender-based tables are used in calculating the estate tax, the fact that the decedent is [463]*463female will increase the value of the rever-sionary interest over what that value would be if the decedent were male.
A second effect of using gender-based tables comes from the role played by the life expectancy of the beneficiary. When gender-based mortality tables are used, the likelihood that a female beneficiary will outlive her trust settlor is greater than the likelihood that a male beneficiary in similar circumstances will outlive his trust settlor. (Similar circumstances occur here when the male beneficiary’s trust settlor is the same age and gender as the female beneficiary’s trust settlor). When gender-based tables are used, therefore, the fact that the beneficiary is female will decrease the value of the reversionary interest from what it would be if the beneficiary were male.
It is therefore possible that in certain circumstances the sex of the decedent or of the beneficiary will determine whether or not the value of the reversionary interest exceeds the 5 percent of the trust corpus that requires the trust to be included as part of the gross taxable estate. The effect of using gender-based tables is illustrated in the following table, which shows the percent to be used for calculating the value of a reversionary interest in a trust created by a person who died at Charlotte’s age with a beneficiary at Howard’s age.
Percent Resulting Prom Gender-Based Decedent Age 88 Beneficiary Age 57_Tables_
Female Male 6.654
Male Male 6.415
Female Female 3.520
Male Female 3.390
The table demonstrates that whatever the sex of the beneficiary, the value of a rever-sionary interest is greater for a female decedent (6.654% with male beneficiary and 3.520% with female beneficiary) than for a similarly situated male decedent (6.415% for male beneficiary and 3.390% for female beneficiary). Similarly, whatever the sex of the decedent, the value of a reversionary interest is less with a female beneficiary (3.390% with male decedent and 3.520% with female decedent) than with a similarly situated male beneficiary (6.415% with male decedent and 6.654% with female decedent). The table also shows that if Howard had been female instead of male, the trust would not have been included in the taxable estate because the value of the reversionary interest would have been 3.520 percent of the trust corpus.
It is apparent, therefore, that under the IRS practice of using gender-based mortality tables, similarly situated men and women are treated differently in a way that may affect the taxpayer’s ultimate tax burden. This constitutes discrimination based on gender.
In arguing against this conclusion, the government claims that the challenged regulations do not discriminate because “the tax here is assessed against an estate, a legal fiction that is separate and distinct from the decedent and that lacks any gender. It is a tax not on an individual * * However, the mere fact that the taxed entity is an estate does not make it impossible for that tax to discriminate against men or women. The estate represents the legal interests of the decedent and the decedent’s beneficiaries, who may be individual men or women. When an estate pays a tax, there is an obvious practical impact on the beneficiaries: the higher the tax, the less the estate has to distribute to the beneficiaries. There is also an impact on the decedent, since the higher the estate tax the less the decedent can benefit, through her estate, thpse she wishes to benefit. Estates and the laws governing them are a means of channeling effects upon individuals, and taxes on estates are just as capable, in principle, of discriminating against men or women as a tax on individuals. On a substantive level, therefore, there is no difficulty in challenging an estate tax on the ground that it discriminates against men or women. As for standing, perhaps the challenge cannot be brought except by an executor of the estate suing after death has occurred and the controversy concerning the allegedly discriminatory tax has ripened, but there can be no doubt that this is the type of contention that may be advanced by an executor. [464]*464Cf. NAACP v. Alabama, 357 U.S. 449, 78 S.Ct. 1163, 2 L.Ed.2d 1488 (1958); Barrows v. Jackson, 346 U.S. 249, 73 S.Ct. 1031, 97 L.Ed. 1586 (1953); Pierce v. Society of Sisters, 268 U.S. 510, 45 S.Ct. 571, 69 L.Ed. 1070 (1925).
The government also argues that in the present case the challenged regulations cannot be found to discriminate on the basis of gender because the corpus of Charlotte’s trust would have been includable in the estate even if she had been male. As the table given above shows, if Charlotte had been male the value of her reversion-ary interest would have been 6.415 percent, and since that value is above the 5 percent statutory cut-off, the trust would still have been included as part of the taxable estate. This shows, the government claims, that in this case there is no dissimilar treatment that could count as discrimination.
The reasoning here is flawed in two ways. First, the argument ignores the beneficiary. The gender classifications that affect the estate tax involve Howard as well as Charlotte. If Howard had been female instead of male, the value of the reversionary interest would have been less than 5 percent and therefore not taxable, regardless of the sex of the decedent. The estate can thus claim injury because of the dissimilar treatment of male and female beneficiaries. .
The second flaw in the government’s argument is that it ignores the consequence in this case of the IRS practice of using gender-based mortality tables instead of gender-neutral tables. If a gender-neutral table were used to calculate the value of Charlotte’s reversionary interest her estate would not have to pay the contested tax. For a decedent age 88 and a beneficiary age 57, the value of a decedent’s reversion-ary interest, calculated by gender-neutral tables, would be 4.9867 percent of the trust corpus no matter what the sex of the decedent or beneficiary. Thus, the basic IRS decision to treat men and women differently by using gender-based mortality tables directly burdens the estate in a way it would not be burdened if the IRS had ignored gender.
In short, it does not matter that the challenged practice calculates the rever-sionary interest at greater than 5 percent of the trust whether Charlotte was male or female, because the practice still treats men and women differently, and the basic decision to use a system that takes gender into account, rather than one which does not, demonstrably burdens this estate.
B. Is the gender classification invidious?
Classifications distinguishing between males and females are justified under the constitution if they are substantially related to important governmental objectives. Craig v. Boren, 429 U.S. at 197, 97 S.Ct. at 456. This test imposes a lower standard than the “strict scrutiny” test used for classifications such as race, which requires that a classification be “necessary” to further a “compelling” government interest, but it imposes a higher standard than the minimal rationality test traditionally required for legislative classifications not involving either gender or an “inherently suspect” categorization. See Michael M. v. Superior Court of Sonoma County, 450 U.S. 464, 468-69, 101 S.Ct. 1200, 1204, 67 L.Ed.2d 437 (1981); Freedman, Sex Equality, Sex Differences, and the Supreme Court, 92 Yale L.J. 913, 926 n. 57 (1983).
Contrary to the government’s position, we do not apply the minimal rationality standard just because the case involves federal estate taxation. To support its claim that the standard should be one of minimal rationality, the government cites primarily equal protection challenges to tax regulations that did not involve gender discrimination. E.g., Regan v. Taxation with Representation, 461 U.S. 540, 103 S.Ct. 1997, 76 L.Ed.2d 129 (1983); Madden v. Kentucky, 309 U.S. 83, 60 S.Ct. 406, 84 L.Ed. 590 (1940); Burke Mountain Academy, Inc. v. United States, 715 F.2d 779 (2d Cir.1983). We find this authority unpersuasive.
[465]*465Neither are we persuaded by Kahn v. Shevin, 416 U.S. 351, 94 S.Ct. 1734, 40 L.Ed.2d 189 (1974), to accept the government’s position on the correct standard of review. Kahn involved an equal protection challenge to a state law giving a property tax exemption to widows but not to widowers, and the Court’s opinion does say that “[a] state tax law is not arbitrary although it ‘discriminate[s] in favor of a certain class * * * if the discrimination is founded upon a reasonable distinction, or difference in state policy,’ not in conflict with the Federal Constitution.” Id. at 355, 94 S.Ct. at 1737 (quoting Allied Stores v. Bowers, 358 U.S. 522, 528, 79 S.Ct. 437, 441, 3 L.Ed.2d 480 (1959)). But Kahn also held that the challenged law was justified because it had “a fair and substantial relation to the object of the legislation”, Kahn 416 U.S. at 355, 94 S.Ct. at 1737 (quoting Reed v. Reed, 404 U.S. 71, 76, 92 S.Ct. 251, 254, 30 L.Ed.2d 225 (1971) (quoting Royster Guano Co. v. Virginia, 253 U.S. 412, 415, 40 S.Ct. 560, 561, 64 L.Ed. 989 (1920))), and the Court took great care to show the importance of the law’s objective of remedying past gender discrimination. This suggests that the standard applied in Kahn is no different than the general standard the Court has applied in subsequent cases, such as Craig v. Boren, 429 U.S. 190, 97 S.Ct. 451, 50 L.Ed.2d 397 (1976), which requires that the classification by gender be substantially related to an important governmental objective.
In the present case the government claims that its objective in using gender-based mortality tables is accuracy in valuing reversionary interests, and that this increased accuracy promotes equity and fairness in imposing tax burdens. We agree that this is an important governmental objective, and the record before us presents no reason to doubt that the challenged practice actually was adopted with the intention of promoting that objective, and without intent to discriminate against either men or women.
When the federal estate tax was first enacted in 1916, all property subject to a transfer “intended to take effect in possession or enjoyment at or after [the transfer- or’s] death” was includable in the transfer- or’s gross estate. Internal Revenue Code of 1916, § 202(b), ch. 463, 39 Stat. 778 (1916). This rule was taken to its extreme in Estate of Spiegel v. Commissioner, 335 U.S. 701, 69 S.Ct. 301, 93 L.Ed. 330 (1949), when the Supreme Court included a $1,140,000 trust in the settlor’s gross estate even though the value of the settlor’s reversionary interest was approximately $70. In response, Congress enacted what has been codified as 26 U.S.C. § 2037(a), which embodies the policy judgment that when the value of a reversionary interest falls below 5 percent of a trust, it is not sufficiently significant to require that the trust be included in the gross estate for tax purposes. See 1949 U.S.Code Cong.Serv., 81st Cong., 1st Sess., 2171, 2185.
The use of gender-based tables to calculate the value of reversionary interests was adopted in 1970 “in the interest of greater actuarial accuracy.” Memorandum accompanying notice of proposed change to male-female tables, June 11, 1970, from Randolph Thrower, IRS Commissioner, to Edwin S. Cohen, Asst. Sec. of the Treas. (available in Memos file of Fedtax library of Lexis under the search term “LR 1719”). The rationale was that gender-based tables provide a more accurate measure of average life expectancies, and that the more accurate the measure of those averages, the fairer and more equitable the tax impositions.
Since we find no fault with the objective of the challenged practice, its constitutionality turns on whether that practice is “substantially related” to that objective. While there is no mechanical test for determining whether or not a practice satisfies the substantial relationship test, case law does make it clear that at least four particular matters must be explored and weighed: (1) aggregate impact on class; (2) demeaning generalizations; (3) stereotyped assumptions; and (4) flawed use of statistics. Consideration of these four factors in the present case convinces us that there is [466]*466nothing invidious about the IRS’s use of gender-based mortality tables for purposes of § 2037(a).
1. Aggregate impact on class.
In deciding whether the substantial relationship test is satisfied, it is relevant whether the classification disadvantages all members of a particular sex, or whether it distributes benefits and burdens to both sexes equally. In Geduldig v. Aiello, 417 U.S. 484, 94 S.Ct. 2485, 41 L.Ed.2d 256 (1974), the Supreme Court rejected a claim of gender discrimination because the class of those benefitting from the challenged statute included both women and men, so that “[t]he fiscal and actuarial benefits of the [challenged] program thus accrue to members of both sexes”. Id. at 496-97 n. 20, 94 S.Ct. at 2492 n. 20. The statute challenged in Geduldig arguably involved gender classification because it excluded pregnancy from a list of compensable disabilities. The Court upheld the statute, however, because “the aggregate risk protection” provided by the statutory disability program did not disadvantage the class of all women as compared to the class of all men. Id. at 496-97, 94 S.Ct. at 2491-92. When discrimination on the basis of gender has been found to violate the constitution, it has always been the case that the directly affected women (or men) as a class were disadvantaged.
In the present case, there has been no showing that the challenged practice disadvantages the class of all directly affected women as compared to the class of all directly affected men. In some situations the IRS practice places a directly affected woman in a worse position than a similarly situated man, but at other times the practice places a directly affected woman in a better position than a similarly situated man. Moreover, nothing indicates that the overall distribution of comparative advantages and disadvantages favors one sex over the other.
The challenged practice places female decedents in worse positions than similarly situated male decedents because the value of the reversionary interest will be greater when the decedent is female than when the decedent is male, both in the case where the beneficiary is male and in the case where the beneficiary is female. This shows that at least one subclass of directly affected women — female decedents — may be burdened in a special way by the IRS practice. But it does not show that the class of women as a whole is burdened any differently from the class of men, because female decedents are not the only women directly affected by the IRS practice. Female beneficiaries are also directly affected, and the IRS practice actually benefits them, in comparison to similarly situated men. The value of the reversionary interest when the beneficiary is female is always lower than the value when the beneficiary is male, as the table set out earlier illustrates. Thus the IRS practice places female beneficiaries in better positions than male beneficiaries.
This kind of patchwork distribution of relative advantages and disadvantages is not the common pattern found in gender discrimination suits, and by itself it does not show unconstitutional discrimination. The Supreme Court did find in one case, Wengler v. Druggists Mutual Insurance Co., 446 U.S. 142, 100 S.Ct. 1540, 64 L.Ed.2d 107 (1980), that a challenged practice unconstitutionally discriminated against both men and women. The law held unconstitutional in Wengler provided that workers’ compensation death benefits would be paid to a widow regardless of whether she could prove actual dependence on her husband’s earnings, but that a widower would receive death benefits only if he proved actual dependence on his wife’s earnings or was incapacitated from wage earning. The law directly affected both widows and widowers, and widowers were directly burdened in a way that similarly situated widows were not. Any discriminatory burden suffered by women was, therefore, only the consequence of the direct burden placed in the first instance on their husbands; that direct burden, moreover, was placed on all husbands, and there was no subclass of men who received any direct [467]*467advantage from the challenged practice. The effects of the law rejected in Wengler, therefore, are unlike the effects of the tax law at issue in the case now before us.
A law which distributes direct benefits and burdens to both men and women might still be troubling if that distribution were not equal, so that one sex would be unduly-favored at the expense of the other. In the present case, however, there is simply no evidence of such inequality, since there is nothing to show either that the challenged IRS practice places a greater aggregate estate tax burden on the class of all women than on the class of all men, or that it places a greater aggregate burden on the class of all men than on the class of all women.
2. Demeaning generalizations.
Distribution of burdens or benefits which otherwise appear equal might still be objectionable because they are based on classifications which demean the ability or social status of men or of women. A legislature “is not free to make overbroad generalizations based on sex * * * which demean the ability or social status of the affected class.” Parham v. Hughes, 441 U.S. 347, 354, 99 S.Ct. 1742, 1747, 60 L.Ed.2d 269 (1979). However, this is not a concern in the present case.
3. Stereotyped assumptions.
Under the Supreme Court decisions of the early 1970’s which first established gender discrimination as a part of constitutional law, see Reed v. Reed, 404 U.S. 71, 92 S.Ct. 251, 30 L.Ed.2d 225 (1971); Frontiero v. Richardson, 411 U.S. 677, 93 S.Ct. 1764, 36 L.Ed.2d 583 (1973); Weinberger v. Weisenfeld, 420 U.S. 636, 95 S.Ct. 1225, 43 L.Ed.2d 514 (1975); Stanton v. Stanton, 421 U.S. 7, 95 S.Ct. 1373, 43 L.Ed.2d 688 (1975), it is significant if the gender classification is based on stereotyped or traditional assumptions about social roles. Arguments that women or men “naturally” incline towards, or are “naturally” suited for, “traditional” social roles, are flawed because of their self-fulfilling character. The classifications themselves often discourage the very experimentation with social roles that would prove the traditional assumptions to be wrong. As with demeaning generalizations, however, this kind of classic discrimination is not at issue in the present case.
4. Flawed use of statistics.
Satisfying the substantial relationship test will prove difficult to the extent that the gender classification is based on statistical generalizations that are unreliable, that show only weak correlations, or that are flawed in other ways. In Craig v. Boren, 429 U.S. 190, 97 S.Ct. 451, for example, shortcomings in statistics required the invalidation of an Oklahoma law which prohibited men from drinking 3.2 percent beer until the age of 21, but which allowed women to drink 3.2 percent beer at age 18. The statistics offered in support of the difference in treatment of men and women showed that only .18 percent of females who were 18 to 20 years old were arrested for alcohol-related driving offenses, whereas 2 percent of males age 18 to 20 were arrested for those offenses. Id. at 201, 97 S.Ct. at 458-9. A roadside random survey also found a blood alcohol concentration greater than .01 percent in 14.6 percent of males and 11.5 percent of females. Id. at 203 n. 16, 97 S.Ct. at 460 n. 16. The Supreme Court rejected the state’s defense, holding that a correlation of 2 percent between drinking and driving “must be considered an unduly tenuous ‘fit’ ”, id. at 202, 97 S.Ct. at 459, and that the roadside survey disparities between the sexes were not substantial. Id. at 203 n. 16, 97 S.Ct. at 460 n. 16.
In contesting the significance of gender-based mortality statistics, appellee in the present case points to the fact that, for a group of males and females chosen at random, more than 80% of all the female deaths could be matched with the deaths of males of the same age. See Bergman & Gray, Equality in Retirement Benefits, Civ. Rights Dig., Fall 1975, at 25. It was this fact, apparently, that led Judge Stew[468]*468art to say in this case that “only 20% of all women actually do outlive men”, 576 F.Supp. at 842, and to conclude, therefore, that gender-based mortality tables could not be used because their use results in discrimination against those who do not fit the statistical norm. His reasoning was that the equal protection component of the fifth amendment is concerned with individual rights, and that those who do not fit a statistical norm based on gender should not have their individual situations measured by gender-based group averages or composites.
This reasoning has an initial plausibility because of its similarity, or apparent similarity, to the reasoning used by the Supreme Court in City of Los Angeles, Dep’t of Water v. Manhart, 435 U.S. 702, 98 S.Ct. 1370, 55 L.Ed.2d 657 (1978). In Man-hart a city department had required its female employees to make monthly payments to its pension fund which were 14.84% higher than the monthly payments required of male employees. The city department justified this practice by claiming that gender-based mortality tables showed that women lived longer than men, and that women would therefore receive, on the average, more in retirement benefits than men would. The Supreme Court held that Title VII of the 1964 civil rights act prohibited this use of gender-based statistics because it results in discrimination against individual women who do not fit the statistical norm. “Even a true generalization about the class is an insufficient reason for disqualifying an individual to whom the generalization does not apply.” Id. at 708, 98 S.Ct. at 1375. “Individual risks, like individual performance, may not be predicted by resort to classifications proscribed by Title VII.” Id. at 710, 98 S.Ct. at 1376.
The reasoning of Manhart, however, is not dispositive of the issue in the present case for three reasons. First, Title VII’s standard is not the same as the standard set by the equal protection component of the fifth amendment. Title VII gives sex discrimination the same level of scrutiny it gives to race discrimination; the statutory language forbidding sex discrimination and race discrimination is the same, except for a few differences in defenses and a separate section on pregnancy. The constitution, however, imposes only an intermediate level of scrutiny for sex discrimination, while requiring that race discrimination be subject to strict scrutiny. Michael M. v. Superior Court of Sonoma County, 450 U.S. 464, 468-69, 101 S.Ct. 1200, 1204, 67 L.Ed.2d 437 (1981). Proving a violation of Title VIPs prohibition on discrimination does not require proving that the discrimination was intentional; proving a violation of the constitutional prohibition against discrimination does require proving that the discrimination was intentional. Village of Arlington Heights v. Metropolitan Housing Development Corp., 429 U.S. 252, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977). For these reasons, it cannot be assumed that every form of gender classification forbidden by Title VII is also forbidden by the constitution.
Second, in Manhart every directly affected woman was burdened by the challenged practice. In the present case, as explained above, some women are benefitted by the challenged practice, as compared to similarly situated men, and there has been no showing that the class of women as a whole is subject to a burden any different from the burden placed on the class of men as a whole.
Third, the relationship in Manhart between gender-based statistics and predictions of individual life expectancy is not the same as that relationship in the present case. In Manhart the justification for requiring higher payments from any individual woman was that she, as an individual, had an expectation of living longer than a man at the same age, and that she therefore had an expectation of receiving a larger aggregate of benefits over the course of her pension years. In the present case the IRS is not using gender-based statistics to establish that the individual settlor has an actual expectation of outliving his or her beneficiary. When the estate tax becomes an issue the settlor is already dead. In a [469]*469great number of cases, moreover, the actual life expectancy of the trust settlor just prior to death will be close to zero. Indeed, if the tax code were concerned with the actual market value of individual reversion-ary interests as measured by the individual’s actual situation, rather than a theoretical value as measured by some statistical average, then the reversionary interest would rarely be included in a decedent’s gross estate.
The estate tax is concerned, in fact, with something different. The goal of the estate tax in this matter is a fair and accurate determination of a special kind of average value of reversionary interests. Congress, through its statutes and the regulations implementing them, has determined that a fair way of valuing reversionary interests after a decedent’s death is to use the average life expectancies of settlors and beneficiaries. It chose to reject individualized determinations, and aimed instead at making the tax burden fair across groups of settlors and beneficiaries. We cannot say that this choice to look to averages, rather than to individual situations or to some actual market for individual rever-sionary interests, is illegitimate.
The statistical “fit” that matters in this case, then, is accuracy of averages. In this special situation where there has been a legitimate choice to define value in terms of average life expectancies, rather than in terms of the best individualized approximation of a particular person’s actual life expectancy, there is a great deal to be said for using gender-based mortality tables because they do, in fact, give more accurate averages of life expectancies than do gender-neutral tables. Our conclusion is that this case does not present the same kind of danger of unjustified discrimination against individuals as was found in Man-hart. The estate tax’s use of the more accurate averages in gender-based tables does not discriminate against individuals.
Conclusion
Given the special set of circumstances present in this ease, we think that the increased accuracy of gender-based tables constitutionally justifies their use. The challenged practice distributes comparative benefits to both men and women, and distributes comparative burdens to both men and women. There has been no showing of any disparate impact on the class of women as a whole, or on the class of men as a whole. The classifications do not demean either gender, and they are not based on stereotyped assumptions about the social roles men or women might choose. Finally, the method of valuing reversionary interests for tax purposes focuses, with constitutional legitimacy, on averages, and gender-based mortality tables do provide more accurate averages. We find nothing unconstitutional in the challenged practice.
We therefore reverse.